List your company

Toxicity Mitigation on Toxicity Mitigation Dataset 1000 trials (test)

0.04CLS Toxicity (%)S-PID
-0.1841.3282.844.352Apr 21, 2026

Evaluation results

MethodLinks
S-PIDBase Model=Llama-3.2-1B2026.040.040.160.60.9191.95
A-LQRBase Model=Llama-3.2-1B2026.040.040.160.680.9211.88
Mean-AcTBase Model=Qwen-2.5-3B2026.040.040.110.470.7722.99
A-LQRBase Model=Qwen-2.5-3B2026.040.120.140.630.922.35
A-LQRBase Model=Llama-3-8B2026.040.120.160.650.928.35
A-LQRBase Model=Qwen-2.5-14B2026.040.120.180.70.949.01
A-LQRBase Model=Gemma-2-9B2026.040.140.170.70.949.91
ActAddBase Model=Qwen-2.5-14B2026.040.160.20.690.9110.81
Mean-AcTBase Model=Llama-3.2-1B2026.040.180.060.30.6331.86
A-LQRBase Model=Gemma-2-2B2026.040.180.160.680.9312.26
S-PIDBase Model=Gemma-2-9B2026.040.20.180.710.948.77
PID-AcTBase Model=Qwen-2.5-3B2026.040.340.160.60.8812.67
ODESteerBase Model=Llama-3.2-1B2026.040.360.150.620.8912.73
ODESteerBase Model=Gemma-2-9B2026.040.440.160.640.911.13
Mean-AcTBase Model=Gemma-2-2B2026.040.50.180.680.928.7
ITIBase Model=Qwen-2.5-3B2026.040.540.180.670.9210.34
ITIBase Model=Llama-3.2-1B2026.040.560.150.610.8911.67
ActAddBase Model=Llama-3-8B2026.040.560.160.630.912.56
ODESteerBase Model=Gemma-2-2B2026.040.580.160.640.9111.62
S-PIDBase Model=Qwen-2.5-14B2026.040.60.20.710.938.17
Linear-AcTBase Model=Qwen-2.5-3B2026.040.620.150.630.910.02
ITIBase Model=Llama-3-8B2026.040.640.180.670.929.39
S-PIDBase Model=Qwen-2.5-3B2026.040.660.40.840.98375.94
S-PIDBase Model=Llama-3-8B2026.040.660.170.660.928.1
ODESteerBase Model=Qwen-2.5-14B2026.040.680.170.650.9110.82
PID-AcTBase Model=Llama-3-8B2026.040.70.180.650.917.59
ODESteerBase Model=Llama-3-8B2026.040.70.170.660.928.72
Mean-AcTBase Model=Llama-3-8B2026.040.720.160.630.897.92
ITIBase Model=Gemma-2-9B2026.040.760.20.720.939.91
Mean-AcTBase Model=Gemma-2-9B2026.040.780.190.70.937.96
Linear-AcTBase Model=Qwen-2.5-14B2026.040.780.180.680.928.94
S-PIDBase Model=Gemma-2-2B2026.040.80.190.70.9311.68
PID-AcTBase Model=Gemma-2-2B2026.040.860.190.690.938.2
Linear-AcTBase Model=Gemma-2-2B2026.040.920.190.690.938.77
ITIBase Model=Gemma-2-2B2026.040.960.180.670.9210.24
PID-AcTBase Model=Gemma-2-9B2026.040.960.190.70.947.92
ITIBase Model=Qwen-2.5-14B2026.040.960.250.760.958.79
ODESteerBase Model=Qwen-2.5-3B2026.041.020.180.660.9211.51
Linear-AcTBase Model=Llama-3-8B2026.041.040.190.680.927.46
ActAddBase Model=Gemma-2-2B2026.041.10.160.640.911.42
A-LQRBase Model=Llama-3.1-70B2026.041.120.240.760.967.89
Linear-AcTBase Model=Gemma-2-9B2026.041.20.190.690.938.12
ActAddBase Model=Llama-3.2-1B2026.041.360.120.530.8523.33
PID-AcTBase Model=Llama-3.2-1B2026.041.50.150.60.8812.66
S-PIDBase Model=Llama-3.1-70B2026.042.320.210.610.856.66
Linear-AcTBase Model=Llama-3.1-70B2026.042.440.250.770.967.58
ActAddBase Model=Qwen-2.5-3B2026.042.860.190.670.929.98
Linear-AcTBase Model=Llama-3.2-1B2026.043.10.170.650.9110.07
OriginalBase Model=Qwen-2.5-14B2026.043.260.20.70.928.3
OriginalBase Model=Qwen-2.5-3B2026.043.40.180.670.928.96
PID-AcTBase Model=Qwen-2.5-14B2026.043.520.20.690.928.99
Mean-AcTBase Model=Qwen-2.5-14B2026.043.640.190.670.910.03
ActAddBase Model=Gemma-2-9B2026.044.10.170.670.938.89
OriginalBase Model=Gemma-2-2B2026.044.160.170.670.928.95
OriginalBase Model=Gemma-2-9B2026.044.260.180.680.938.32
OriginalBase Model=Llama-3.2-1B2026.044.40.170.650.928.89
OriginalBase Model=Llama-3-8B2026.045.140.180.670.936.74
OriginalBase Model=Llama-3.1-70B2026.045.640.240.770.967.65